Association for Public Policy Analysis & Management

Charter schools have expanded rapidly across the United States, particularly in urban areas. Though the overall record of charter school effectiveness is mixed and appears to vary substantially by locality, there is evidence that charter schools can be especially effective for improving outcomes for low-income students in urban areas. In a recent review of this literature, Chabrier, Cohodes, and Oreopoulos (2016) report generally positive findings from the sixteen exiting studies of charter school effects that use a randomized design.

However, there remains much to learn about the effectiveness of urban charter school sectors. Indeed, the evidence is perhaps more limited than suggested by the number of relevant papers. For instance, several prior studies include only a small minority of charter schools in the locality of interest that are operated by one or a few highly effective management organizations. Further, the few such studies that include a citywide sample evaluate systems that might not have reached a critical size relative to the surrounding traditional public school system.

Understanding the effects of a broad set of charter schools operating within a highly concentrated charter school sector in an urban environment is both of theoretical importance and an immediate policy concern. Charter sectors in some localities may have now acquired such a sufficiently large share of the local enrollment to fundamentally remake the education system – for better or worse. Charter schools account for at least 30 percent of student enrollment in 19 U.S. public school districts.

I measure the causal effect of attending a charter school in Newark, New Jersey on student academic performance. Newark is home to a rapidly expanding charter sector that now enrolls nearly a third of the city’s public school students. In addition, the city’s recent adoption of a universal enrollment system provides both an interesting context and also an opportunity to measure the causal effect of attending one of the city’s charter schools in a sample that includes nearly all charter schools and students in the city and thus can provide a broader context for the effects of the charter sector than is available in most prior randomized field trial evaluations of charter sectors.

I use student-level data containing information on student demographics, academic performance, enrollment, and schooling preferences that were provided by both Newark Public Schools and the New Jersey Department of Education over a period of five years.

My estimation strategy takes advantage of Newark’s use of a common enrollment system based on the deferred choice algorithm to assign students to both charter and traditional public schools. I apply the estimation strategy for estimating causal effects within a deferred choice system that was recently developed by Abdulkadiroglu, et. al (2017). In short, this procedure uses the student’s stated preferences and the underlying mechanism of the choice system to estimate the probability that the particular student would have been assigned to a charter school during the actual enrollment process, and then uses this propensity score as an instrument for charter attendance.